On Wednesday 08 September 2010 13:15:31 Dennis Nezic wrote:
> On Wed, 8 Sep 2010 11:07:07 +0100, Mark wrote:
> > 
> > >>>> You shouldn't run into problems doing a direct copy, assuming you
> > >>>> change the four ports mentioned in freenet.ini. You will simply
> > >>>> have multiple instances of freenet running, each accessible via
> > >>>> it's own http port, each accessing it's own udp data pipes/local
> > >>>> sockets. I think.
> > >>> thank you for getting back to me.
> > >>> If I understand you correctly, if I copy the "freenet"
> > >>> directory/subdirectories from machine A to another machine in a
> > >>> different geographical location, say machine B, I don't need to do
> > >>> anything but start the service (run.sh start). All of my
> > >>> configuration files, trusted friends, etc. will be duplicated from
> > >>> machine A to machine B. Then, a new KEY will be issued by machine
> > >>> B that I can use to make friends with machine A. Correct?
> > >> Oh ... hrmm ... right, each node has it's "identity key" ...
> > >> stored in the opennet-OPENNETPORT and node-DARKNETPORT files. I'm
> > >> not sure what would happen if two nodes with the same "identities"
> > >> run at the same time. (They will, of course, but I'm not sure
> > >> what, if any, collisions will happen. I don't think I want to find
> > >> out either :b.) Perhaps (my wild guess) if you delete these two
> > >> files they will be regenerated?
> > >
> > > There is yet another problem.  The cache (both of them) will
> > > contain information from the originating Machine A. It seems, there
> > > is no easy way to clone.
> > >
> > I'm currently working on a bartPE install of Freenet. I have a working
> > system now (I'll post that in the thread I started about this),
> > however I hit the same type of problems.
> > 
> > What I did in the end was to install freenet on a clean PC, run the
> > wizard once and then copy all the installation files - but I emptied
> > the datastore directory. When Freenet is started it recognises that
> > the datastore is empty and corrects itself. It seems I had to keep the
> > caches (node.db4o(.crypt), temp-PORTNUMBER and
> > persistent-temp-PORTNUMBER) otherwise freenet bombs out with a
> > complaint that these didn't exist although I suspect there's a
> > configuration I could set in freenet.ini that stops that error
> > (anyone know what it is?)
> 
> Hmmm -- I regularly delete those caches, for various reasons, without
> problems. 

Such as?

> How long ago did you have that problem? (Maybe it was 
> corrected since then.)
> 
> Also, I still think there will be a problem with cloned nodes having
> the same identity. At least with ARKs (which are node-unique USKs that
> hold IP information, to help with NATs, I assume) -- each of the
> clones, I believe, will be inserting different (IP) information to the
> same USK. And I suspect other conflicts will arise too. (If you have
> two clones connected to you as peers, will your node see them as
> different nodes even though they have the same "identity (keys)"? If
> not (perhaps they will be regarded as multiple addresses to the same
> node, similar to 127.0.0.1 and 192.168.0.1), since they will each
> eventually have different "locations" in the freenet topology, their
> contribution to freenet will not be avaiable to you.) That's my
> handwaving, at least.

On Monday 13 September 2010 22:54:31 Dennis Nezic wrote:
> On Mon, 13 Sep 2010 16:51:15 -0400, Uriel Carrasquilla wrote:
> > If I had a P2P with only 3 nodes that I own, then I would not have
> > any exposures. If I have a darknet, is it through some trust that
> > security can be achieved? What makes darknet so much more secure than
> > opennet?
> 
> Yes -- you actually (hopefully) know and trust each of your peers,
> unlike opennet strangers. I believe that is the only significant
> difference. (To infiltrate a/the darknet, physical surveillance /
> kidnapping / bribing / torture is necessary.)

No. This is an important point. You do not need to trust your peers with your 
life. You just need to trust them more than you do the peers automatically 
assigned on opennet. Given that attackers can pretty much connect to whomever 
they want to connect to, they can given sufficient resources connect to 
everyone, they can cycle through the whole network a group at a time, and so 
on, your family, coworkers, regular Settlers of Catan partners, and even online 
acquaintances, are probably safer than opennet - unless your primary purpose in 
finding them was to find people to connect to on Freenet (as with asking total 
strangers on an IRC channel), in which case they are probably the enemy.

The point really is about increasing the cost of the attack. Of course it is 
possible for a determined state to make everyone's friends betray everyone. But 
it's way more expensive than the simple low budget electronic attacks that are 
possible on opennet.

BTW electronic surveillance e.g. remote root compromise is also possible. But 
even that is more costly because it is hop by hop and there is a nonzero chance 
of getting found out.

> > > -- you actually (hopefully) know and trust each of your peers,
> > > unlike opennet strangers. 
> > May be I have watched too many 007 movies, but what if one of your
> > trusted peers is actually a double agent?
> 
> That's a good question. Maybe someone more knowlegeable can help flesh
> out the details, but I recall reading a while back that it's possible
> for peers to know what is in each other's datastores/caches? (Via a
> timing attack... faster retrievals imply something exists?) Although
> one still has plausible deniability so long as you have at least one
> non-compromised peer, so I'm not sure how meaningful this would be.

It is possible to determine what is in your peers' datastores, however it is 
not useful because for the first few hops we don't cache anything (except in 
the client-cache, which can't be read remotely and is encrypted).

Unfortunately, "plausible deniability" is a bit of a myth in most cases. As 
long as there are enough samples (whether these can be predicted in advance 
matters for mobile attacker source tracing but not for more brute force attacks 
such as connecting to everyone), it is possible to get a fairly good idea of 
whether the peer is the source of the requests.

On Wednesday 15 September 2010 04:58:38 Uriel Carrasquilla wrote:
> 
> > To move in DarkNet you actually have to go and talk to a person... something
> > like "Hi, do you mind introducing me to some of your friends?" which may 
> > work
> > only sometimes.
> 
> It seems that we are pushing technology to the point that a breakdown to 
> remain anonymous could be 
> our human condition more than a technical one.  But we have been wrong before 
> in regards to technology.
> So can we say that the anonymity problem in P2P networks is solved?

Definitely not. Freenet is vastly less secure than it ultimately needs to be. 
Even the darknet is vulnerable to a powerful denial of service attack that we 
haven't dealt with yet. Opennet is hideously insecure, and even on darknet it 
is quite possible to attack your neighbours or to gradually move towards the 
source of a stream of data. We have lots of ideas on how to improve the 
situation. Tor may in fact provide better anonymity *if* it is available where 
you are - but it does have its own weaknesses, for example it can be blocked 
rather too easily, and it is probably *slightly* more vulnerable to traffic 
analysis.

Attachment: signature.asc
Description: This is a digitally signed message part.

_______________________________________________
Support mailing list
Support@freenetproject.org
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://emu.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:support-requ...@freenetproject.org?subject=unsubscribe

Reply via email to